Awesome
IEEE ICCV-2023
Monte Carlo Linear Clustering with Single-Point Supervision is Enough for Infrared Small Target Detection, Boyang Li, Chao Xiao, Longguang Wang, and Yingqian Wang [Paper]
Algorithm Introduction
We propose a Monte Carlo linear clustering (MCLC) method to acheive high-performance infrared small target segmentation with only single point annotation. Experimental results on several datasets (e.g., NUAA, NUDT, IRSTD-1K, and NUDT-Sea) verify the effectiveness of our method. The contribution of this paper are as follows:
-
We propose the first single point supervised method to achieve SIRST detection.
-
Inspired by the inherent characteristic of SIRST, a simple yet effective linear clustering approach with random noise-guided Monte Carlo regularization is proposed to coarsely extract and further refine the candidate target region.
-
Ablation study reveals that pixel-level labels are not necessary for SIRST detection while single-point supervision is strong enough.
Citation
If you find the code useful, please consider citing our paper using the following BibTeX entry.
@inproceedings{li2023monte,
title={Monte Carlo linear clustering with single-point supervision is enough for infrared small target detection},
author={Li, Boyang and Wang, Yingqian and Wang, Longguang and Zhang, Fei and Liu, Ting and Lin, Zaiping and An, Wei and Guo, Yulan},
booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
pages={1009--1019},
year={2023}
}
Prerequisite
-
Tested on Ubuntu 16.04, with Python 3.7, PyTorch 1.7, Torchvision 0.8.1, CUDA 11.1, and 1x NVIDIA 3090 and also
-
Tested on Windows 10 , with Python 3.6, PyTorch 1.1, Torchvision 0.3.0, CUDA 10.0, and 1x NVIDIA 1080Ti.
-
The NUDT-SIRST download dir (Extraction Code: nudt)
Usage
1. Preprocessing.
Replace the example image in ../dataset/final_dataset_final/NUAA-SIRST/images with the original full images.
Replace the example mask in ../dataset/final_dataset_final/NUAA-SIRST/masks with the original full masks.
cd final_size_Prior_code
python generate_single_point_Prior.py
2. Pseudo Label Generation.
cd final_size_Prior_code
python Baseline_NUAA.py
3. Retrain SIRST Detection Network (e.g, DNANet) with the Generated Label.
SIRST Label Geneartion Software
Based on the algorithm in the above paper, we developed a single-point annotation software. Download dir